Meta-Heuristic and Non-Meta-Heuristic Energy-Efficient Load Balancing Algorithms in Cloud Computing

Introduction

In IT industry, the demand of energy is constantly rising. Recent study shows that in US, 10% of the total commercial electricity is consumed by ICT and that is 20% in Germany (Koomey et al, 2010). The study also projects that the power requirement of data centres in US will be increased drastically from 60TWh/y in 2005 to 250TWh/y by 2017 (Aebischer et al, 2015).These projections and predictions are generating a requirement towards use of sustainable and energy efficient mechanisms and approaches in ICT sector (Procaccianti et al, 2015). Cloud computing (CC) is a most recent computing paradigm (Buyya et al, 2009) which is widely used in IT services based on pay-as-you-go model (Kord et al, 2013). Some of the major challenges lie in the underlying domain is: data privacy, legal regulations, utility and risk management, on demand provisioning and security (Zapata et al, 2015).However, energy efficiency has turned out to be a main apprehension especially in cloud data centers (Liu et al, 2011). In US, data centres put away a fraction of 1/15 of the whole commercial electricity use in 2006, whose monetary value is approximately of 4.5 billion per year (Dabbagh et al, 2015). As a consequence the proprietors of data centre are very much serious to adopt naive ways for energy saving strategies which can minimize their operational costs. Along with this, it is reported in (Pettey et al, 2007). that the ICT sector contributes 2% of total carbon emission which is environmental concern and is an alarming issue and a great concern for everyone (Borah et al, 2015). It has been investigated that the top 500 High performance computers consume 20 MW of energy which is equivalent to a small thermal power plant’s power consumption (Filiposka et al, 2016). The energy charges and carbon footprints are surely going to increase in coming days as data centres are expected to raise radically in terms of volume and count in order to meet the current service needs. As an outcome the prime objective of the entire fraternity of industry, academia, and government agencies in present days are towards searching solutions and methods which can lessen energy consumption in all aspects of computing by using standard energy conservation techniques.

A broad classification of energy management schemes in CC environment is shown in Fig. 1. The energy management schemes can be largely classified into two groups as 1) Static Energy Management (SEM) and 2) Dynamic Energy Management (DEM) (Beloglazov et al, 2011). The SEM schemes are based on methods which use optimization of design time at different levels. In circuit level the optimized methods try to lessen power consumption caused due to switch flipping in logic gates and transistors by using optimized number and size of transistors. In logic level, the techniques are aimed to be employed to minimize the power consumption due to switching activity of logic gates and transistors by using appropriate complex gate design and minimizing the number of transistors (Dasgupta et al, 2013). But, only by optimal hardware design and use is not sufficient to achieve better performance in terms of energy consumption. The overall performance can be enhanced only if at both hardware and software level, optimized design techniques will be used.